Multimodal sensing and AI
Fusing multiple sensing modalities (touch, sound, vision) to estimate the physical properties of objects (no prior knowledge)
Human-to-robot object handover
Handling dynamic, safety-critical transfers of unknown objects and the unpredictability of human movement
Open benchmark ecosystem
Building a community tha improves, shares, reproduces protocols, ML models, and data, and compares on our leaderboards the
Datasets
Multimodal recordings of people (or a robot) manipulating objects, and real and mixed-reality images of handheld containers.
Browse datasetsSoftware
Handover pipelines, real-to-simulation framework, robotic control algorithms, and perception models for object recognition.
Explore softwareBenchmark & protocol
Evaluation toolkit and leaderboards to compare methods on a standard benchmarking protocol designed for the task.
View protocol